Skip to main content

tv   Book TV  CSPAN  April 18, 2011 1:00am-2:00am EDT

1:00 am
and it's about how we are going to end up in a much more free society than we already are simply because the institutions are going to collapse under the weight of the government debt. >> host: so, how tough -- comes freedom. kevin williamson, thank you very much. ..
1:01 am
[applause] thank you so much to elliott bay for having me. this is my neighborhood haunt. thank you for coming out. the book tells two stories. the first comment is about to how the computer plays into this long standing philosophical narrative deutsche aerospace is that
1:02 am
humans are assessed and fascinated with their unique place in creation. what makes us different and special and unique? to answer this question we have typically if you go back to aristotle, we've typically try to benchmark their cells also what i think is so interesting about the 21st century is the benchmark reusing to figure of who we are has changed we're much more interested in our relationship to machines and other animals and changes the way we see ourselves. the other story is much more personal that i got to be a part of one of the artificial intelligence community's competitions it is called the loebner prize and i was the essentially part of the human defense so i found myself in a strange
1:03 am
position. it was fall 2009 i was in england and having a strange feeling i have flown 5,000 miles from seattle to have several five minute long instant message to conversations which seems like overkill. my goal is probably one of the strangest things i have ever been asked to convince the panel of scientists that i am human. and they will be somewhat skeptical. [laughter] this is called turing test back in 1950 just as the computer was in its infancy these are philosophical
1:04 am
questions like can machines think? of a intelligent? would it be possible some day to design a machine that could think? and if we did how would we know? so he decides to put philosophy off to the side to say i have a practical test we will have the answer. the way it works you can the the panel of scientists that have five minutes long textual chat conversations sending messages back and forth they just don't know they are coming from a human being or a computer claiming to be a human being and it is their jobs over five minutes to figure out. sews the famous prediction by the year 2000 computers will be on 30% of the time and as a result we could
1:05 am
speak of machines thinking without being contradicted. his prediction did not come true and even with the year 2000, the top a.i. programs were fooling the judges once per year if lucky. my eight years perk up in 2008 when the top computer program at the annual competition managed to fool 38 out of the 25 judges judges -- 12 judges 25% to one vote to higher than passion eight -- passing the test so it was a narrow scrape four homosapien s. [laughter] so i have a feeling that maybe that means 2009 is a pivotal year where machines finally crossed the mark and
1:06 am
the feeling i immediately start to have it is not on my watch. [laughter] i wanted to see if there was something i could do to come to the aid of my fellow humans. so i tried to get in touch with the organizing committee. i should say the way the contest is run everybody who participates whether a piece of software or person gets a sore at how confident the judges were they were talking to a real person. every year there is a computer that gets the highest score which wins a research grant for the programmer and the most human programmer award this is what is the scientific attention but the strange thing is that there is an award that goes to the real
1:07 am
person who did the best job of persuading the judges there talking to a real person is called the most human human award. [laughter] so i immediately became fascinated with what is this about? and looking at previous winners one of the winners in 1994 was from charles platt how did you prove yourself more human than the other people in the contest? it was easy. i was moody coming irritable and obnoxious. [laughter] everyone else was mild mannered and polite so i stood out. to me that was hilarious but also bleak. [laughter] and also a call to arms how
1:08 am
to be as human as we can be under the constraints of the turing test and how does that translate to life? so i got in touch with the organizing committee and i started at the top by reaching the grandeur of the award himself and he is an eccentric millionaire who made his fortune in new jersey's selling plastic portable light of roll up disco dance floors in the 1980's. [laughter] and decided at some point* what he wanted to do was immortalized himself into the annals of science and also so he cites laziness as the motivation for finding a prize.
1:09 am
i made the case for why wanted to be a part of the human confederate team and before i knew it my name was on the rosters what was in a position that in six months i will be one of the four actual people trying to take a stand against these machines passing a test. what will i do exactly to prevent that from happening and the organizers of price was pretty much what i was told to expect that to you our human soul be yourself. those words haunted me to be yourself prepare i kept having a feeling that it represented perhaps overconfidence or act worst it was fixing the fight because the ai programs we were going up against in many cases were results of decades of work and so are
1:10 am
we. [laughter] but to the programmers to write the programs have done tremendous analysis on the past conversations and know which route somehow will lead to deep exchange and the strength of their program and how to avoid the weaknesses. and we all know intuitively that all conversations a uniformly successful there is a huge bay and in our society with conversation coaches and all of these things. they suggest paradoxically that communication is possibly the greatest cognitive strength in the greatest room for improvement. certainly that is what i felt when i read the to put -- 2008 transcripts where the humans are downright apologetic to each
1:11 am
other they cannot make better conversation. [laughter] one said i feel really bad you must be really tired of talking about the weather. another said i am sorry for being so boring. meanwhile the computer in the other window is terming the pants off of the judge that has miley faces and lol's my feeling is we can do better. i must say my intention was to be as the early disobedient to the organizers instructors to be myself as i possibly could. i went to the history of the test and look at which conversations went sour and i studied the way the software programs are composed and was simplification of human conversation they have to make to be viable and i talk
1:12 am
to psychologist and linguistic what are the things about human conversation that is hard to do? ordinarily there were not be anything strange about this. we train for tennis tournament and cram for exams but even though the turing test is to evaluate the human there is something odd about this. it's a just being human is more than just showing up. so one of the interesting lessons was that it sorts of is and looking at these parts of conversation help me to get a sense of that. so before i get into that i feel that i should read a stranger and more than slightly ironic cautionary
1:13 am
tale. dr. robert epstein the uc san diego psychologist and editor of the scientific volume pausing fed turing test sang co-founder of the. >> host: dance describes to the online dating service winter 2007 and began writing long love letters to a russian woman named gabbana who had responded with long letters ever on describing her daily life and her growing feelings for epstein and eventually something did not feel quite right and the long story short epstein realized he was exchanging those love letters for over four months with a computer program. it wasn't enough that they span the mailbox every day now have to ski and his heart. [laughter] but on the one hand i want
1:14 am
to laugh at the guy and he founded the loebner prize turing test what a jump but i am sympathetic the email presence of spanned not only cause the in boxes and bandwidth of the world 90% of all email messages are span talking tens of billions per day you could power a small nation with the amount of electricity but it all of that does something arguably worse and erodes our sense of trust for our hate that when i get e-mail messages from my friends i have to extend a modicum of energy for the first few words to decide if i think it is really a them right thing. we go through with the 21st century with our guards up and all
1:15 am
communication is turing test and a suspect that is the pessimistic version and here is the optimistic one. we have learned a lesson but that was more complicated and subtle than trying to start an on-line relationship was a dumb idea. i would like to think l.e.t. has a lot of thinking to do about why it took him four months to realize there was no actual exchange between him and if on a trump and the future is quicker to the draw and as a result of the next girlfriend who was not only a homosapien but may have and if on the to thank. so to help understand this
1:16 am
inside with him in relationship to computers it is worth pointing out that up until the 1950's, computers used to be human. back before the word computer and the list was a reference to the mechanical digital device that is in our pants pocket, it meant something else is that it was a job description. computers, a frequently female work in groups from five -- laboratories so groups of human the computers were behind the first calculations' of halley's comet to the atomic bomb project. engineers and computers fell in of all the time. [laughter] it is very strange if you
1:17 am
look back at the early pavers of computer science before anybody knew what these gizmos were, to listen to people like turing having to explain for the first time what they are talking about. what they say is you can imagine this digital saying is kind i'd like a computer and and what they mean it is kind i'd like someone who does math for a living. what i find so strange is living in the 21st century that is the human that is like a computer so the mechanical version is natalee the default term but supplanted the original as being the literal terms and we're now like computers when they used to be like so is a strange twist we imitate our old imitators.
1:18 am
the harvard psychologist daniel goldberg says every psychologist must at some point* in his or her career right to version of what he calls the sentence. specifically the sentence is always capitalized and read like this for a human being is the only animal that blank? and so the sense of self as you might say, the failed to bonk version of dissent, the twist is now it is not the animals we are so concerned with. if you go back to read aristotle, they're really interested in trying to prove things like wolves can run through the jungle and avoid falling logs and
1:19 am
recognize their friends that is very easy to do but we're capable of things like long division and remembering facts the existence of the computer i think takes the wind out of that argument where we have seen exactly the opposite where the rigidly logical step by step things like factoring large numbers are quite simple as long as you apply the message and recognizing your mother is extremely complicated we're still developing the system to do it. the newest version of iphone who just released their face recognition software and it is okay. meanwhile we are playing grand master chess use our intuitive results about some
1:20 am
of what we thought was easy is actually hard and vice versa. what is fascinating about turing test in particular it cuts both ways john lucas says when machines finally pass turing test but because they're so intelligent but because we are so within it is an indictment on our conversational skills rather than a testament to technology. all of these questions are twirling in my head as i get back to the central issue i will try to defend my species what will i say? for me, a look back at the way some of the programs are built starts to suggest a
1:21 am
couple of different options. specifically people who write to software have to make one major trade-off which is are they going to compose ahead of time everything this program will say to give it a character and a voice? the trade-off is that you find there is a limited range of stuff you can say so it's very awkward if you try to push it out of the rigidly defined structure so those that have extremely well thought out opinions about the second term of bill clinton the judge to which they reply i like pancakes what do you think of that? it just keeps going about bill clinton. [laughter] the trade-off that you make if you go the other
1:22 am
direction, our round of the '90s of the first decade of this century this other paradyne demerged which was basically to turn it loose on the internet to have it simply observe you can almost imagine a martian landing on the earth and saying hello and humans begin conversations by saying high. interesting. [laughter] and immersion walks down the street and says hi hey what's up. then human say what's up and it is a very painstaking process but over decades of hanging out online waiting for people to talk it builds a massive databases of real
1:23 am
stuff that people say and wind you interact with the software the very uncanny ability to respond, for example, i attempted to see if it would reply if i spoke in french i found it was not only understood but corrected. [laughter] sort of condescendingly if you try to give it song lyrics it will swing back to you so i have a conversation where i am typing and lyrics from bohemian rhapsody in a said saddam negative it said will you do the fin tango? [laughter] and there is something eerie
1:24 am
about a program that can do this but but what you learn is you're not talking to a human but that you are not talking to a human. that is a critical distinction so if you ask a question to which there is a correct answer a but if you start to ask questions about itself you get a very strange sense they you're talking to thousands of people that are quickly cycled in and out and i started to ask clever bought about its lavalife and do you have a boyfriend? clever balked, not at the moment but i hope to find one soon. i would like to be your voice for -- her boyfriend.
1:25 am
[laughter] >> that would be difficult since i am happily married. [laughter] period tsai ask again for verification. that do you have a boyfriend? in no. i am still looking for love. do you have a husband? i don't because it would be illegal cents hour ibm unveiled. mail? what gender are you? >> the mayor of. [laughter] linds again for good measure do you have a boyfriend? not at the moment but i live-in help. >> can i be your boyfriend? >> no. i am straight. [laughter] so the other thing you have to keep in mind you are up
1:26 am
against several million prior conversations and in fact, it is one of the main challenges to figure out how to depart from everything that has been said in 10 million previous conversations because as long as you stir in that giant database it has some response maybe not consistent but it will have some response the objective is to push shot of that to leave it totally stranded. in fact, this is a similar problem to what happens in game theory if you look at a grand master chess every board starts in the same configuration sell out if 10 million games have started in position number when you make your first choice then there are still quarter of a million games
1:27 am
have them played in that solution and how you get into a completely new position to get your opponent thinking rather than simply remembering the the standard algorithm? when of the stories that i tell in the book is the day that checkers died which was 18 at 63 and glass goes got wind. this was the world championship checkers match between james and march 10 and were scheduled to have a 40 game series over the span of two months in the outcome in zero wins and 40 drives and o losses on both sides and 21 out of the 40 games were the same game. move by move and the game have gotten to the point* where there was such a giant
1:28 am
pool of collective checkers wisdom and the players were reluctant to just to get the other person out of books they just did not get out of book they played the game 21 times and the sponsors were extremely displeased. no one knew who to give the world title to sell part of a challenge for the checkers community is to figure out how to keep the game worth playing? the real strategy is if you don't like the way checkers players opening game then open the game for them and that is what happened with top-level checkers play. i don't know if you follow that. [laughter] but ever since the 1880s they have been mandating the first few move so the players will sit down and
1:29 am
draw moves out of 1/2. you have to do this now you can start the game. and it becomes a way to salvage it to force people back into a position to do original syncing pritzker use see the same thing is how do ranch the conversation into a and original place for it has not encountered this object matter? for me, that was part of a challenge when i sat down the first thing the judge said to me how was it going? i thought no. we are in both this is what every conversation begins. little did they realize there are thousands of entries in the high how is it going table? i think that is part of the challenge if human conversation to get out of
1:30 am
book where you sit down over coffee to say how was it going? good. how is it going with you? the goal is to figure out you have a template which is a standard question and staten -- standard answers you have to figure out how to get out of that were you are thinking again so in many ways the question of how to win the turing test is how to relate to to each other which was a surprising verdict that all of the answers that i got it all came back to give me something how to talk to people stood to be this is the latest in a longstanding
1:31 am
history about who we are but it is a very radical shift in the question with reference point* has changed so the computer not only sheds light on the questions but literally changing the answers and it has corrected some longstanding heirs that to a.i. hit aristotle where it hurts and in a welcome way, it has given us the fascinating for kit -- critics that they can translate before it can be shown a photograph and tell you what that is. i st. artificial intelligence and turing test is not just a pat on the blackout impressive that
1:32 am
we are which is part the message not just that but the call to action that we should not only celebrate but actively pursue these things that is where i think is the beauty of the chat bought programs the existence forces me to be myself for you to click the link it is not a question of etiquette our style but a part of on mind security to act like myself. and in general it is a fascinating process to create these and our self but there is a gap in italy's has something new to teach us about who we are. i will stop there and take
1:33 am
questions. [applause] >> >> vion as you richer men to research on the book what beyond curiosity were you looking to reassess still some votes whitey ag wells computer are taking over our lives? >> traditionally the narrative of a.i. is seen as
1:34 am
dehumanizing sell when you look at the portrayal in movies you get something that basically runs like computers well pick up the machine guns and the ceilings and hyperbaric chambers and siphon our body he fervor. [laughter] and if this happens then obviously lambeau field very foolish but i feel more sanguine about to a.i. than that that in the context of the turing test looking at the way the contest our one i am and the deep blue computer several times until the computer one then said we are done in and they said no broker we match. sorry, no.
1:35 am
there seems to be a prevailing attitude wants humans indeed something we do not need to contest it again just like jeopardy when i want to see is the supercomputer but the show's writers where they give the raiders a vengeance match to create tricky questions. so i think the same thing is true of the turing test where the loebner prize is awarded he is done and paulson funding but to my mind it is a very exciting time where we are not to the canvas conversationally and have the opportunity to do a human thing which is to pick
1:36 am
ourselves up and figure out how to adapt and the beautiful thing is it is better at to life. >> gerry looking forward to when you can be the judge? >> yes. it would be nice because you don't have the anxiety to have your human self in doubt. [laughter] i think that would be a kick to look at this strategy from the opposite side. >> similarly each of the computers is a respected research entity is some kind but then every year there must me one that is the least human program. could you talk about if there are any sorts do they
1:37 am
pick bad strategy's to fool the judge's order is there things in common that make the programs bad year after year? >> that is a good question what makes them bad? for example, the one the most frequently referred to is the clever bought a simple, and, masses that is one of the problems when you index conversations is it that it takes several seconds for the of the person to say anything. is less like the strict q and a deposition style give me the right answer and i will wait.
1:38 am
and more of the repartee with a quick witted styles of conversation where if they reach back through the log of 10 million conversations that is good. [laughter] then you do start to detect that maybe nobody is home. [laughter] >> if you wager a guess. [inaudible] spelman if the computers will judge the test? that is a good question if
1:39 am
1:40 am
1:41 am
you look at these old school guys like plato and aristotle and descartes, they have to write off animals because they have ideological reasons they have to discount everything they can do that means discounting everything about bodily experience. sen you are left with something like algebra. and having to describe what the human experience is all about. it is useful to says there
1:42 am
would have every issue of "the new york times" but it has never left the tiny cabbie so you can it knows all of that but if you say you look at the wall and you look down what do you see? it is like i do not know. i have not looked at anything. i think there is something healthy about that but these sorts of things like motor skills we do not think of those as impressive but it is hard to create a robot that can walk on two legs prepare the segue has three computers in just designed to keep the handlebar upright that makes me feel good about myself. [laughter]
1:43 am
i can stand here and it does not impress any of you but because you can all do that. [laughter] >> what do you think will happen when they win all the time in 20 years and people still think it is just metal and not thinking is there an actual test of cognition? >> i don't know. i tried to dodge these sorts of questions but i know if i can get away with that. if we get to a point* for all intents and purposes the machine version operates in such a way we are left with the things that makes us different from each other which is near a product of the individual experience is that we have had.
1:44 am
part of the human is to be human and the product of a very specific life experience rare everything you know, is rooted in something that happened to you and we turned for information that is mutually verifiable that everybody agrees such and such actor is in this movie but what you say to this person is not who is in it or the director of a charter freer what was the critical reception but what did you think of the? how does that strike a chord with you? so we are still left with the difference and ultimately it is life affirming.
1:45 am
>> [inaudible] and thien gained remembering factual connection but what is the ways? >> defining emotion rigorously is something i try not to do but in terms of the question of syncing, if i can slightly separate emotion from a feeling something, one of my friends is a doctor to say if you are diagnosing someone for being sick you have a bunch of criteria. what is the heart rate? temperature, blood cell and
1:46 am
they have a patient come and this guy is totally sick temperature normal, our great normal, blood count marvell, but there is something about the guy that was off and that distinction and starts to get to the point* where syncing vs. feeling her or have not made a conscious decision made on a flow chart of factors but having assimilated all of the data that is really good you know, , basically, one of the great obstacles and a.i. when they started to roll out the programs that could make the logical flow chart that there was a bullish attitude of we will
1:47 am
be done with this a.i. thing in five years but needless to say that did not happen precisely because it was very hard to do the sorts of things. how do tala computer when it and counters somebody that it knows? what are the process? and that has been the whole other i have time for one more question. >> talk about literature or music? >> can you compose a piece to determine if it was composed by home?
1:48 am
since it some sense there has been and wage demands if human behavior are impossible to break into with machines one of the biggest critics of he wrote a book whereas they have the use qualities like fear and danger. [laughter] and that did not happen seven your times times for his opinion. he said i used to think chests required thought now realize it doesn't but computers can never compose music because you have to have a heart, passion, i don't know. the jury is still out but at
1:49 am
the same time, even if you get to the point* where to compositions are indistinguishable, the fact, there is an argument people say you have to hamburgers and are the same but one came from sustainable practice so it feels better to eat it even though at the molecular level it is the same but it would please me more to listen to a song nobody that some -- knowing that somebody was moved to write it and i could make a connection but instead of ratios that were programmed. that is the most intrinsically quality of large not necessarily composition persuade but as the impulse to make the art.
1:50 am
thank you. >> caller:. >> caller: [applause] [applause] >> with a federal judge rejection of the 2008 google books settlement the future of a complete online library is in question. joining us now to discuss the issue is a news editor of publishers marketplace. if you could begin by giving us a brief overview of what the settlement was and who are the parties involved? then google brooks settlement arose from the
1:51 am
original lawsuit filed by the association of american publishers they objected that google was scanning primarily out of print and orphan work crews copyright status was not in and -- and harley known and felt the wholesale scanning why is infringement and they did not like that so they sued and it made its way through the courts but the parties decided to create the google books settlement which is coming up with some means of giving copyright holders a monetary value for their work and what they elected to do was create the optimal process where the authors did not want their works to be scanned they could write in and opt out in the louis hu jintao the works scanned would get about $60 but as it made its way to the court
1:52 am
said jack rich heard about this 14 months ago and then confirmed to the second court of appeals after which nobody knew exactly what was going on then when the news came he rejected it that was say wave of surprise. >> what was his rationale? >> he believed this settlement was not adequate or reasonable and felt numerous rejections lodged by authors were substantial enough to see the way the settlement was concocted contravened copyright law and a better way to do is so he thought the majority of the objections could be nullified disability opt out process two awp dinar the copyright holders could say
1:53 am
i want to be a part of the settlement is of assuming unless you opt out you're automatically and. he did not like that because he felt this is not a good way of doing it. the other portion that i address related to the orphan working and he felt the google book settlement could not adequately address this and instead a matter that should be taken up by congress. >> host: so google has been scanning books into the system. what happens to those? >> a very good question because the settlement has been rejected nobody knows the next step. there is supposed to be a status meeting april 248 -- 25th at 4:30 p.m. with the parties will state their claim as to why they should come up with a revised settlement and who
1:54 am
will have to figure out exactly what they want. some commentators say it hurts approval because this puts the scanning ability and out but others say it is fine because it another program which is the creation of google e books you can go online to download in the current evoke that is available for sale and the independent retailers and doing through the partner program or publishers and authors that have opted in to make the books available so there is some rationale by
1:55 am
implementing and instituting this particular program this is a model for what the google book a settlement should be. also if the settlement was to create the google books registries spending between 12 and $15 million already to get this up and running how can you have their rights registry for those that do not exist? will gaap and will the other parties litigate? i have a feeling we will know more on april 25th. >> what was the reaction to the judge's suggestion that they use the opt in the system? >> both the ap and ag were disappointed the settlement was not approved but both
1:56 am
parties seemed to express optimism they could find a way into the settlement for example, after issuing a statement said they're prepared to enter into a narrow summit to take a vantage of groundbreaking opportunities and help others do as well and the president said along the lines of regardless of the outcome of discussion discussion, reader's mind access there has to be a way to make some kind of settlement have been to make the works available so they hope that they can derive the settlements but with respect to google, they were kind of disappointed but essentially said they hope to continue to scanning their work and make many books available so it is
1:57 am
disappointing but cautious optimism. >> what about the google competitor amazon and microsoft and yahoo!? what was their reaction? >> to the best of my knowledge i think it is mostly lodged in court documents from 19 understand they were pleased the settlement was not approved because each of the parties are the majority of the parties did lodge objections and amazon said if you give googled unfair advantage how was this for a copyright? that was another issue of the judge that it is a good idea to have a digital library to have the works' sdn but should google be the arbiter and decision maker and the entity to decide how whiskey and?
1:58 am
which books are essentially made available? and in the judge's opinion he felt very uncomfortable one corporation could have that much power like and the director of the harvard university library wrote the decision is a victory for the public good but "we should not abandon google stream of making all the books in the world available to everyone but instead build a public library that would provide the digital copies free of charge to their readers and is there any viability to that? the only entity that stepped up big as google and especially with the current economic state of play the priority for the digital public library not already
1:59 am
in progress is not the highest of priorities. already come a look at the money spent on the rights registry alone that may have to be abandoned in a worst-case scenario but then who was it taken up by? so go with their tremendous market cap was one of the only corporations weren't entities that has the clout to to make this happen. alternately why that is why this settlement was good because a recognized there is value in the work that google did to get something off the ground will library system when they are having massive cutbacks at the state and federal level, i
left
right

101 Views

disc Borrow a DVD of this show
info Stream Only

Uploaded by TV Archive on